Summary

IOSM CLI is a terminal-native AI engineering runtime designed for professional developers, offering controlled, measurable, and repeatable execution directly against codebases. It addresses the need for systematic engineering work by orchestrating parallel agents, managing complex tasks, and tracking metrics over time, moving beyond conversational AI to a robust execution environment. The primary benefit is enabling audited, benchmarkable development cycles with enhanced operational safety.

How It Works

The core of IOSM CLI is its terminal-native AI runtime, optimized for controlled engineering execution on filesystems and shells. It employs the IOSM (Improve, Optimize, Shrink, Modularize) methodology, transforming ad-hoc refactoring into a reproducible, measurable process tracked by six canonical metrics. Key architectural choices include a Policy Engine v2 for deterministic permissions, filesystem checkpointing for reliable rollbacks, and a flexible MCP (Model Control Protocol) server architecture for extensibility via custom tools and agents. This approach prioritizes auditability, repeatability, and benchmarking over conversational AI.

Quick Start & Requirements

Installation is straightforward via npm (npm install -g iosm-cli) or npx (npx iosm-cli). Prerequisites include Node.js version 20.6.0 or higher and at least one authenticated AI model provider (e.g., Anthropic, OpenAI, Gemini, Groq). For enhanced code searching capabilities, optional dependencies like ripgrep, fd, ast-grep, comby, jq, yq, and semgrep are recommended, particularly for large repositories. The project provides extensive documentation, including a getting started guide.

Highlighted Details

• IOSM Methodology: A systematic engineering improvement framework (Improve, Optimize, Shrink, Modularize) with six core metrics (semantic, logic, performance, simplicity, modularity, flow) and quality gates for measurable progress.
• Complex Change Workflow: Supports a controlled progression from /contract (defining scope) to /singular (analyzing options) and /swarm (deterministic execution with locks, gates, retries, and checkpoints).
• Policy Engine v2 & Security: Implements deterministic layered permission resolution, session/turn-scoped approvals, and a security baseline for package/extension sources with trust ledger and integrity verification.
• Filesystem Checkpointing: Integrated snapshot/restore functionality ensures deterministic rollback and audit trails for command execution.
• Extensibility: Features MCP servers and TypeScript extensions, allowing custom tools, slash commands, UI components, and provider adapters.

Maintenance & Community

The project is actively developed, with version 0.3.13 indicating ongoing updates. The primary author is Emil Rokossovskiy. No specific community channels (like Discord or Slack) or sponsorship details are mentioned in the provided README.

Licensing & Compatibility

The project is licensed under the MIT license, which is generally permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

As of version 0.3.13, the tool is under active development. Suspend behavior is disabled by default on Windows due to potential signal path issues, though unsupported environments now recover more gracefully. While designed for robustness, certain terminal environments might still present challenges with hotkeys or suspend functionality. The reliance on external AI model providers means performance and availability are subject to those services.